1 / 39

Microsimulation and public policy. Issues and prospects

Microsimulation and public policy. Issues and prospects. Wellington, Feb 2012 martin@spielauer.ca. What is Microsimulation? Microsimulation at Statistics Canada Examples Conclusions. Outline. Simulation = creation and the use of models pursuing a purpose: Exploration Prediction

aldan
Download Presentation

Microsimulation and public policy. Issues and prospects

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Microsimulation and public policy. Issues and prospects Wellington, Feb 2012 martin@spielauer.ca

  2. What is Microsimulation? Microsimulation at Statistics Canada Examples Conclusions Outline

  3. Simulation = creation and the use of models pursuing a purpose: Exploration Prediction Problem-solving Training Explanation and theory building Raising consciousness Many policy questions can only be answered using simulation What is simulation?

  4. Social science microsimulation: Computer-simulation of a society or economy in which the population is represented by a large sample of its individual members and their behaviors What is Microsimulation?

  5. If individuals are different and differences are given importance No single representative agent Study and projections of distributions If behaviors are more stable or better understood on micro level Non-linear tax and benefit rules Composition effects If individual histories matter Non-Markov processes: agents have a memory Pension projection: contribution history When does MS make sense?

  6. Tax-Benefit analysis Pension Education Population projections Health Spread of diseases Care Personnel planning Where is it used?

  7. Increase (policy) relevance of data collected by STC Data from a single source often can’t answer all questions MS allows (and requires) the integration of various data into a coherent platform Makes data more accessible Confidential micro-data might be releasable when integrated into a synthetic database Data quality Detection of data gaps and inconsistencies Internal data analysis and modeling expertise together with close client relations feed back Why MS at Statistics Canada?

  8. Long history going back to 1984 Large variety of modes (15+; e.g.:) Static tax benefit model: SPSD/M Large “multi-purpose” socio-economic model: LifePaths Population projection model: Demosim A growing family of health models: Pohem Modgenmodeling technology and programming language “Common language” at STC Worldwide use Provision of modeling support, collaborations & consultancies Academia – Governments – International Organizations Microsimulation at Statistics Canada

  9. Time Line

  10. Modeling Division Socio-economic models: LifePaths, SPSD/M Modeling technologies: Modgen Internal and external modeling support, model prototypes Health Analysis Division Fast growing family of health and disease models Various academic and other collaborations Demography Division Demographic projection models Human Resources Development Division Personnel models Who at STC develops MS models?

  11. STATIC MICROSIMULATIONExample: SPSD/M

  12. Model for analysis of tax-transfer changes in a given year Federal & provincial transfers and taxes Static accounting model Includes tax and transfer rules from 1991 to 2015 Tax and transfer model is controlled by over 1600 parameters Includes commodity taxes & consumption Starting database: synthetic database (~200,000 individuals) created using 4 datasets STATIC MICROSIMULATIONExample: SPSD/M

  13. SPSD/M

  14. Accessibility: Releasable synthetic data from 4 sources Quality: Testable! Relevance: many users! STC provides tool and training Clients run own scenarios Users across political spectrum SPSD/M: Data integration

  15. Technology:Modgen programming language

  16. Dynamic MS introduced 1957 Guy Orcutt: “A new type of socio-economic system” Impressive progress: Hardware Statistical software Data: Longitudinal; administrative Programming Reinventing the wheel? Technology 1956 IBM 305 RAMAC First computer with a disk drive - fifty 24" discs – total 4.4 MB total; $35,000/year

  17. Modgen is a generic MS programming language Continuous time and discrete time Interacting and non-interacting populations It is free Fast; large populations; multithreading Efficient: No advanced programming skills required User-friendly: visual interface and detailed documentation. Powerful tabulation: e.g. coefficients of variation for tables; Visualization of individual life courses Modgen technology

  18. Modgen: User interface

  19. Common language for MS at STC: more flexible team Analysts can develop and maintain models themselves Efficiency in development of models Avoid re-inventing the wheel Shared high standards Less errors Fast creation of prototypes Vehicle for teaching Modgen: Conclusions

  20. Population projectionsExample: Demosim

  21. Microsimulation and Demography • Demographic research showcases trends in social sciences and the transformative potential of microsimulation • From structure to processes • From macro to micro towards multi-level integration • Traditional Macro projections limited to few variables. Miss need for more detailed projections; poor micro foundation • MS can replicate macro models which can be stepwise refined accounting for heterogeneity

  22. Population projections showcase one of the fundamental trade-offs in microsimulation: detail versus prediction power Detail leads to specification randomness: too many variables produce noise Simplicity leads to misspecification errors: too few variables, i.e., too simple models Microsimulation allows scaling of models by theoretical considerations Microsimulation and Demography

  23. General: Rapid increase in the ethno-cultural diversity Demand for detailed projections going beyond age and sex Institutional: Funded by external federal departments Results Released 2011 Follow-up studies, e.g. for Aboriginal Population Demosim: Context

  24. Starting population is a 20% sample of the census (7 Mill) Age Sex Place of residence Place of birth Generation status Visible minority group Mother tongue Aboriginal identity & Registered Indian Status Religious denomination Highest level of schooling Labour force participation Demosim – Base Population & Variables

  25. Many models follow a 2 step approach Base risks: equivalent macro models Relative risks: detailed accounting for individual characteristics Fertility Mortality Emigration Labour force participation Education (indirect) Allows to combine robust with detailed data sources Supports scenario building Demosim - “Typical modeling strategy”

  26. Demographic microsimulation projection meets demand for more detailed projections while leaving models transparent Demonstration that MS models can be developed fast with modest input of resources Confidence in results: aggregate outcomes resemble macro projections (and differences are explainable) Scenarios created by clients Very high media coverage Demosim: Conclusions

  27. Large multi-purpose dynamic modelsExample: LifePaths

  28. Attractive features Virtual world: Test and fine-tune new policies Detail: Policy rules at any level of detail; distributions Longitudinal perspective: Sustainability issues Policy maker’s perspective

  29. Simulates a large sample of detailed individual life courses which together represent the Canadian population Dynamic – large - multi-purpose Family context: simulation of spouses, children, grandchildren Continuous time model Synthetic starting population Historic depth: first actors born 1871 Creates a detailed, virtual world for the analysis of policies with a longitudinal component Example: LifePaths

  30. Family demographics Education Health Employment Income and earning Home ownership Social insurance Benefits Taxes Public pension plans Senior benefits (OAS, SPA, GIS) Private pension plans LifePaths – Behaviours & Systems

  31. The content of LifePaths is largely embodied in its behavioral equations estimated from micro data Challenge: both longitudinal and cross-sectional consistency LifePaths makes use of a multitude of data sources. LifePaths can be viewed as a data integration exercise LifePaths – Data

  32. Motivation: concerns about the “replacement adequacy” of the retirement income system for future retirees Why LifePaths: Provides detailed distributional results Complete lifetimes observable: detailed earning histories; flexibility in measuring pre-retirement earnings, e.g. best 10 years Allows exploration of “what-if” questions LifePaths application example: Replacement adequacy of retirement income

  33. Traditional “typical case” analysis Starts work at age 18 Earns the average wage every year Stops working at age 65 C/QPP provides a pension that replaces 25% of earnings Example: Replacement adequacy of retirement income

  34. LifePaths analysis: realistic careers Example: Replacement adequacy of retirement income

  35. Distributional analysis Example: Replacement adequacy of retirement income

  36. Cohort analysis Example: Replacement adequacy of retirement income Source: Kevin D. Moore, William Robson, AlexandreLaurin (2010) Canada’s Looming Retirement Challenge; C.D. Howe Institute Commentary Pension Papers

  37. What-if behavioral analysis Example: Replacement adequacy of retirement income

  38. Complex “general purpose” models come at a price but are very flexible for a wide field of policy applications Policy needs ebb and flow; balancing updates and core development with client’s applications is a difficult task. Many spin-offs (which have to be valued) Expertise; “training lab” Readiness: base for new applications, fast response to new policy demands Contribution to quality and relevance of data LifePaths: Benefits & Lessons

  39. There are many reasons for expecting a boom in MS Many policy questions require MS MS is a logical next step complementing data analysis MS will probably replace many cell-based models:population projection, actuarial models, disease models Technologies like Modgen make MS accessible Microsimulation is still under-used and has to find its way into the standard toolbox of social scientists and into classrooms Students pick up the ideas and develop skills fast Microsimulation has high data demands. Statistical offices are a logical collaborator in microsimulation projects; In return MS can enhance the relevance, quality, and accessibility of data Conclusions

More Related